Integrating SMOTE-Tomek Balancing with Sparrow Search Optimized DBN for Secured and Energy-Aware IoT Networks

Abstract

The exponential expansion of the Internet of Things (IoT) has created an interconnected world characterized by unprecedented convenience and automation, but it has also introduced major challenges in terms of security and energy efficiency. The diversity of IoT devices, the heterogeneity of data, and their constrained resources make intrusion detection and energy optimization critical research areas. This paper proposes an intelligent and hybrid framework that integrates the Synthetic Minority Oversampling Technique combined with Tomek Links (SMOTE-Tomek) for data balancing and a Sparrow Search Algorithm (SSA) optimized Deep Belief Network (DBN) for intrusion detection and energy-aware decision-making in IoT networks. The proposed model mitigates class imbalance in IoT traffic datasets, enhances feature learning through deep hierarchical representations, and optimizes DBN parameters using the bio-inspired SSA. The hybridization of SMOTE-Tomek with SSA-DBN significantly improves model robustness against minority attacks and minimizes false alarms while ensuring low computational overhead suitable for resource-limited IoT environments. Extensive simulations performed on benchmark IoT datasets demonstrate the superior performance of the proposed approach in terms of accuracy, precision, recall, F1-score, and energy efficiency compared to state-of-the-art techniques.